Claude Haiku 5.5

Fast, cheap small model for agent sub-tasks

🧠 LLMs & APIs NEW THIS WEEK provisional score

/ 100

Anthropic's small model released Oct 7, 2026 for high-volume, cost-sensitive work like summaries, sub-agents and browser use.

You run many agent sub-tasks and the token bill hurts.

You need maximum reasoning depth on hard problems.

Score breakdown

Job fit25/30

Ease / time-to-first-result13/15

Value14/15

Momentum13/15

Openness & portability6/10

Trust & maturity10/10

Builder/gamer bonus5/5

Initial editorial estimate from public info; not yet hands-on tested. Provisional (new release, under 30 days). How we score →

Pricing: Pay-per-token via the Claude API (see vendor pricing section)
Unverified. Unverified: editorial summary. We couldn't confirm this on an official pricing page (JS-rendered, no public pricing, or usage-based) as of Oct 18, 2026.
Free tier: No (unverified)
Platforms: API
Open source: No
Launched: Oct 7, 2026
Official URL: https://www.anthropic.com/claude-haiku-5-5
URL verified: Oct 18, 2026 (curl HTTP GET)
Last reviewed: Oct 18, 2026 by agent
Sources: https://www.anthropic.com/claude-haiku-5-5